a flow adaptive aerodynamic probe concept for turbomachinery
TRANSCRIPT
Measurement Science and Technology
A flow adaptive aerodynamic probe concept forturbomachineryTo cite this article: C Lenherr et al 2007 Meas. Sci. Technol. 18 2599
View the article online for updates and enhancements.
Related contentTime-resolved flow measurements withfast-response aerodynamicprobes inturbomachines
-
Time-resolved entropy measurementsusing a fast response entropy probe
-
An optical backscatter probe for timeresolved droplet measurements inturbomachines
-
Recent citationsQuasi Two-Dimensional Flow-AdaptiveAlgorithm for Pneumatic ProbeMeasurementsChristian Bartsch et al
-
One-Dimensional Flow-AdaptiveMeasurement Grid Algorithm forPneumatic Probe MeasurementsChristian Bartsch et al
-
Design Considerations for Axial SteamTurbine Rotor Inlet Cavity Volume andLength ScaleKonstantinos G. Barmpalias et al
-
This content was downloaded from IP address 82.130.73.41 on 17/08/2020 at 15:29
IOP PUBLISHING MEASUREMENT SCIENCE AND TECHNOLOGY
Meas. Sci. Technol. 18 (2007) 2599–2608 doi:10.1088/0957-0233/18/8/035
A flow adaptive aerodynamic probeconcept for turbomachineryC Lenherr, A I Kalfas and R S Abhari
Turbomachinery Laboratory (LSM), Swiss Federal Institute of Technology, ETH Zurich,CH-8092 Zurich, Switzerland
E-mail: [email protected]
Received 6 September 2006, in final form 30 April 2007Published 11 July 2007Online at stacks.iop.org/MST/18/2599
AbstractA flow adaptive 2D traversing algorithm is developed and demonstrated withmeasurements in a large axial turbine facility. This novel approach is suitedfor pneumatic probe and fast response aerodynamic probe measurements.The implementation of the algorithm is fully automated and requires aminimal input, such as blade count and hub and tip diameters, at set-up. Thealgorithm automatically selects measurement points, such as shear flows,secondary flows, wakes through user-defined detection functions, and addsadditional measurement points; therefore higher measurement fidelity inthese regions is obtained compared to a traditional measurement method.The flow adaptive 2D traversing algorithm can resolve the overall flow fieldwith 75% fewer measurement points compared to a uniform a measurementgrid. This reduction in measurement points results in a measurement timeusing the flow adaptive algorithm that is 81% quicker than on a uniformmeasurement grid, without loss of measurement accuracy.
Keywords: adaptive traversing, aerodynamic probe measurement,turbomachinery
(Some figures in this article are in colour only in the electronic version)
Nomenclature
Cpt total pressure coefficient Ptot−Pstat,exit
Ptot,inlet−Pstat,exit(–)
M Mach number (–)Patm atmospheric pressure (Pa)Pstat static pressure (Pa)Ptot total pressure (Pa)α flow yaw angle (deg)β flow pitch angle (deg)d distance (m)v velocity vector (m s−1)
Subscripts
i index, counterx axial directionθ circumferentialr radial
Abbreviations
D dimensionalDF detection functionFP flow property valueNL number of loopsNP number of points/loopNT total number of possible measurement pointsNA actual number of measurement pointsPG pressure gradient sectionPM minimal pressure sectionPS blade pressure sideSS blade suction sideTCx test case no xYA yaw angle section
1. Introduction
The measurement, analysis and prediction of detailed flowfieldfeatures, such as wakes and vortices, are major issues
0957-0233/07/082599+10$30.00 © 2007 IOP Publishing Ltd Printed in the UK 2599
C Lenherr et al
in the development of modern turbomachinery designs.Furthermore, an understanding of the complex physicalprocesses associated with these flow phenomena is requiredto improve the efficiency and performance of turbomachinery.In the past two decades computational fluid dynamic methods(CFD) have played an increasingly important role in the designof modern turbomachinery design, especially for reasonsrelated to development costs. However, CFD methods arestill limited by available computing power in particular whenunsteady flows must be resolved. There are also a number ofissues related, for example, to turbulence or subgrid modelling,numerical accuracy and or boundary conditions that remainunresolved. Thus the role of experimental measurementtechniques shall remain important for the foreseeable future.
Sieverding et al [1] provide a recent overviewof measurement techniques for the unsteady flowsin turbomachinery applications. Due to the harshenvironment and limited accessibility of turbomachinery, pointmeasurements derived from intrusive probes are the bulk oftechniques used in industrial applications. In the industrialsetting, a substantial amount of time, money and effort isexpended in setting up, conducting, reducing and analysingthe large volumes of data that derive from point measurements.There is therefore great interest in the reduction of the requiredtime for turbomachinery tests, as these reduce developmentcosts and maintain competitiveness [2].
We present here a novel flow adaptive traversing algorithmthat has been used with both pneumatic and fast responseprobes [3]. This algorithm requires minimal user input atthe outset, thus reducing the set-up time for a measurementcampaign. During the measurements, in an automated manner,the algorithm detects and measures flow features such aswakes and regions of secondary flows, such as vorticalstructures. This automated adaptation of flow measurementregions is important as regions in which measurements must beperformed are not always intuitive. For example, in blade rowclocking the flow phenomena of interest move with respectto the stationary frame of reference; this movement is notalways proportional to the clocking angle [4]. In this case,the flow adaptive algorithm yields measurements at pointsrelated to the actual flow field, and thus the result with differentclocking angles can be compared in an unbiased fashion. Thenumber of measurement points on the flow-adapted grid issubstantially reduced compared to a uniform grid, without aloss of measurement accuracy. Thus the measurement timeand the time for data reduction are reduced; also there aresavings in the electronic space required to store the data.
The format of the paper is as follows. First, theflow adaptive traversing algorithm is described. Then theinstrumentation and axial turbine facility that are used todemonstrate the algorithm are presented. Measurements ofthe total pressure coefficient and yaw angle on uniform andflow-adapted measurement grids at the second stator exit arethen discussed. The paper then concludes with a summary ofsignificant contributions of this new development.
2. Flow adaptive traversing algorithm
The flow adaptive algorithm is designed to first detectpoints/areas of interest and then in an automated manner, add
Figure 1. Illustration of new enhancement points added around apoint of interest.
more measurement points. The adaptation is accomplishedin both the radial and circumferential directions in themeasurement plane (figure 1).
Three sequential steps—pre-processing, main processingand post-processing—constitute the flow adaptive algorithm.
2.1. Pre-processing
The first step in the flow adaptive method is to define an initialuniform grid of measurement points. This initial uniform gridmust cover the whole measurement domain with an optimumnumber of points such that the measurement time is short andno important flow details are missed. Subsequent flow-adaptedgrids are based on this initial uniform grid. Algorithms thatstart from a uniform start grid to generate an adapted grid havebeen applied in adaptive mesh techniques for CFD [5]; thusin this regard the present approach is novel for measurementtechniques but not new in fluid dynamics.
The optimum number of points to be used on the initialuniform grid was experimentally examined in the 2D axialmeasurement plane, figure 1, of the two-stage axial researchturbine that is described below. This optimum number ofpoints is a function of the geometrical parameters of themeasurement plane, and thus will vary from one facility toanother. For the present facility on the order of 102 (that isapproximately 10 × 10 points in the circumferential and radialdirections, respectively) were found to be optimal.
Typical spanwise profiles of the mass-averaged totalpressure coefficient and yaw angle are shown in figure 2.
The profiles are shown for two grids, an initial uniformgrid and a fine uniform grid. The initial uniform grid has 8 ×11 points and the latter grid has 45 × 59 points. This secondgrid is twice as dense in both radial and circumferentialdirections than a typical measurement grid that is used inthe present facility without a flow adaptive measurementtechnique. In figure 2 the per cent difference between theprofiles on the two grids is shown along the second x-axis.Overall the agreement between the profiles is very good, withthe exception of the measured total pressure at the tip. Inthis case, the relatively coarse initial uniform grid does notcapture the tip leakage effects. However, when we considerthat the initial uniform grid has 30 times fewer measurementpoints (88 compared with 2655) than the fine uniform grid,
2600
A flow adaptive aerodynamic probe concept for turbomachinery
Spanwise distributions of circumferentially mass averaged total pressure coefficients and deviation
between the results
Spanwise distributions of circumferentially mass averaged yaw angles and deviation
between the results
Legend:
Figure 2. Comparison of spanwise distributions ofcircumferentially mass-averaged Cpt and yaw angles on a fineuniform grid and initial uniform grid. The second x-axis shows thedeviation between the two measurements.
the 1%–2.7% differences in the total pressure coefficient and1.2%–2.7% differences in the yaw angle are very small.
After the initial uniform grid is specified, the pre-processing step is concluded with the definition of a finestuniform grid. Not all the points of this second grid are used;rather this second grid is used as the basis for an interpolatedmeasurement grid when regions of interest are detected inthe main processing step of the 2D adaptive algorithm. Thedesired measurement time and details of the probe geometryare amongst the parameters used, together with the initialuniform grid, to define this finest uniform grid.
2.2. Main processing
The refinement to a flow-adapted grid and measurements onthis adapted grid are done during the main processing step.The refinement is based on user-defined detection criteriathat identify regions of interest in the measurement plane.
In the present work detection criteria are specified, and thusthe main processing step has three phases. In each phase adetection criterion is applied, the grid is refined and a seriesof measurements are made. In order to automate the mainprocessing, the number of times to repeat a phase (termedhere, loops), the number of points to be measured in each loopand termination criteria are specified a priori.
The termination criteria are based on user-specifiedminimum and maximum distances and the distances between ameasurement point of interest and its neighbouring measuredpoints. If the distance between the point of interest and apreviously measured neighbouring point is not less than theuser-specified minimum distance, then a new measurementpoint is inserted midway between the point of interest andthe previously measured points. If the distance exceeds auser-specified maximum distance, then the newly insertedmeasurement point is placed at a specified distance fromthe point of interest. This procedure ensures that the newmeasurement points are within the area of interest; ourexperience shows that the overall quality of measurementson the flow-adapted grid is then better. The user-specifiedminimum and maximum distances are defined in terms of thepositioning uncertainties of the probe traversing system andthe diameter of the probe head.
The insertion of new measurement points results inan unstructured grid. However it is computationally moreefficient to perform the data acquisition tasks on a structuredgrid using an object-oriented programming language. Thusafter each refinement, the resulting unstructured grid isinterpolated onto a structured utility grid. This structured gridhas the same grid point spacing as that of the finest uniformgrid. A distance weighed mean averaging procedure is usedfor the interpolation.
As described above, the adaptation of the grid andmeasurements are done in a series of phases. In each phase,a detection criterion is used as the basis for the adaptation.In the present work, the axial turbine flow, that is examined,is characterized by wakes, which are shed from the rotor andstator blades, and vortices that are generated by the passageand leakage flows. Thus three detection functions, minimumpressure (PM), pressure gradient (PG) and yaw angle (YA), areemployed. The pressure-related functions provide for accuratemeasurements in the wake, since the total pressure in the wakeis lower than that of its surrounding flow.
Kalfas et al [6] and Binder et al [7] have shown thatthe yaw angle is a reliable indicator to identify both wakesand vortices, especially in the passage downstream of a stator.Thus the sequence of detection, adaptation and measurementphases, figure 3, is PM, PG and finally YA.
In the PM phase, the adaptation and measurements areperformed if local minima of the measured total pressuresare found. In the PG and YA phases, the adaptation andmeasurements are performed if the detection functions exceeda threshold value. The detection functions for the pressuregradients are
DFi (r) =(
∂Ptot
∂r
)2
i
(1)
DFi (θ) =(
∂Ptot
∂θ
)2
i
. (2)
2601
C Lenherr et al
Figure 3. Flow chart of an adaptation and measurement phase.
And for the yaw angle
DF1i =(
∂α
∂r
)2
i
·(
∂α
∂θ
)2
i
(3)
DF2i =(
∂2α
∂r2
)2
i
·(
∂2α
∂θ2
)2
i
. (4)
The threshold value is given as
threshold = 1
m·
m∑i=1
DFi, (5)
where DFi is a detection function given in equations (1)–(4).It is pertinent to point out that the detection functions involvesquares of the first and second derivatives. This results insmoother distributions of the detection functions and improvedidentification of the points of interest.
2.3. Post-processing
The adaptation algorithm is initiated on the initial uniformstarting grid, and then is applied on successively refinedgrids as areas of interest are identified and measured. Ifa uniformly spaced grid were used for the measurements,then the measurement data would be stored in matrices thatare easily post-processed. However as the 2D flow adaptivealgorithm results in measurements that are clustered in areasof interest, the resultant measurement data are not amenable toefficient post-processing. Thus the measurements on the flow-adapted grid are interpolated back onto a uniform grid in orderto facilitate the post-processing. This interpolation is done intwo steps. First, a 1D cubic spline is used to uniformly spacepoints along the boundary of the measurement area. Then theflow parameters at each uniformly spaced point (x0, y0) areevaluated as a distance weighted average of the four closestmeasurement points. For example, for the yaw angle:
α(x0,y0) =α1d1
+ α2d2
+ α3d3
+ α4d4∑4
i=11di
. (6)
3. Experimental validation of the technique
3.1. Instrumentation
Two probes, a pneumatic cobra-shaped five-hole probe [8],figure 4, and a fast response aerodynamic probe (FRAP)
Figure 4. Cobra-shaped five-hole probe.
(a) (b)
Figure 5. (a) Yaw and pitch angle convention; (b) fast responseaerodynamic probe.
[9], figure 5, are used to evaluate the flow adaptationalgorithm. Gossweiler et al [10] and Johansen et al [11]discuss the calibration procedures for the aforementionedprobes. Traditional calibration techniques are limited to smallflow angles [12], but a theoretical basis for extending the rangeof flow angles, beyond those used in the calibration, has beengiven by Pisasale et al [13]. The pneumatic probe has acalibration range of ±12◦ in the yaw angle and ±30◦ in thepitch angle (figure 5). The diameter of the probe head is0.9 mm and the tip has a slanted pyramid shape (figure 4). Thefive-hole probe yields steady measurements of α, β, Ptot, Pstat
and M.Unsteady measurements are derived from the FRAP
(figure 5(b), [10, 14–18]). This probe incorporates a single,temperature-compensated sensor that is operated using aWheatstone bridge. The diameter of the cylindrical probeis 1.8 mm. The pressure tap is located 1.8 mm from theprobe tip and has an angle of 0◦ with respect to the x-axis(figure 5(a)).
The sign convention is that a positive yaw angle is in thedirection of the rotor sense of rotation (y-axis), and a positivepitch angle is directed towards the blade tip (z-axis).
3.2. Facility
The 2D adaptive flow concept was applied in the axial turbineresearch facility ‘LISA’ [19]. This is a large (max. power400 kW), low speed (Mach numbers 0.1–0.4) facility that canaccommodate up to two axial turbine stages. The primarycharacteristics of the turbine used in this study are given intable 1.
The present measurements are made in a traversing planethat is located downstream of the second stator row (figure 6).
2602
A flow adaptive aerodynamic probe concept for turbomachinery
Stator 2
Rotor 1 Rotor 2
Stator 1
Probe Traversing Planes
Inlet Flow
nd
Figure 6. Measurement position after the second stator.
Table 1. Characteristic turbine parameters.
Parameter Value
Rotational speed 2500 RPMPressure ratio 1.34Aspect ratio (Span/Ax. Chord) 1.8Blade count (Rotor/Stator) 42/42Outer tip diameter 0.8 mMass flow 10.26 kg s−1
Table 2. Test case 1 (TC1).
Phase NL NP NT NA
Initial grid 1 88 88 88PM 3 50 150 118PG 2 40 80 64YA 4 50 200 84
Table 3. Test case 2 (TC2).
Phase NL NP NT NA
Initial grid 1 88 88 88PM 4 50 200 145PG 4 40 160 133YA 4 40 160 147
3.3. Test matrix
The baseline grid used to evaluate the 2D flow adaptivealgorithm is a uniform grid with dimensions of 23 ×60 points in the circumferential and radial directions,respectively. Two test cases, TC1 and TC2, with flow-adaptedgrids are examined. The salient features of these grids aresummarized in tables 2 and 3. The primary differences in thetwo test cases are the number of loops in the PM, PG andYA phases of the main processing step, and the number ofmeasurement points in each of these loops.
The total numbers of possible measurement points are 518and 608 for the test cases TC1 and TC2, respectively. However,due to the termination criteria based on distances that arediscussed in section 2.2, the actual numbers of measurementpoints are 354 and 513, respectively, compared to 1380 pointson the baseline uniform grid. In the subsequent sectioncontour plots and circumferentially mass averaged line plotsof the measured total pressure coefficients and yaw angles arepresented for the two test cases to show the application ofthe 2D flow adaptive algorithm. In the contour plots, circlesymbols that show the location of the measurement points
Figure 7. Cpt distribution for test case TC1.
Figure 8. Cpt distribution for test case TC2.
are superposed on the flowfield contours. Although otherflow variables could be shown on the plots, the total pressureand yaw angle are presented since they are also used in thedetection functions.
3.4. Discussion of results
Figures 7 and 8 show contours of the total pressure coefficientsat the end of the measurement series for test cases TC1 andTC2, respectively. Since a five-hole probe and FRAP are usedfor the measurements, a safety clearance must be maintainedclose to the hub. Additionally near the rotor tip, there is aleakage flow that results in yaw angles that are outside thecalibration range of the probe. Thus in figures 7 and 8, andin the subsequent contour plots, the regions in the range of0–5% span and 99–100% span are shown as blanked out whitezones.
A comparison of test case TC1, figure 7, to test case TC2,figure 8, shows that although the former has approximately30% fewer measurement points, there are no significantdifferences between the two cases.
2603
C Lenherr et al
Figure 9. Yaw angle α (deg) distribution for test case TC1.
Figure 10. Yaw angle α (deg) distribution for test case TC2.
This is confirmed in figure 11 where the spanwisedistributions of the circumferentially mass-averaged totalpressure coefficient and yaw angle are shown, and comparedto those of the reference grid. It can be seen from the circlesymbols in figures 7 and 8 that the 2D flow adaptive algorithmresults in more finely resolved measurements in the regions oflow total pressure that are associated with the wake. For testcase TC2, table 3, more loops involving the pressure-relatedcriteria are performed in comparison to the test case TC1,table 2. Thus a comparison of figures 7 and 8 also showsthat test case TC2 has a higher density of measurement pointsin the wake and its surrounding flow. This higher density ofmeasurement points is also evident when the yaw angles arecompared in figures 9 and 10.
Figures 9 and 10 also show that there is a relatively highdensity of measurement points in the outer 25% span of themeasurement plane. In this region the tip leakage vortex resultsin relatively high gradients of the yaw angle.
A close-up view of the measured total pressure and yawangles is shown in figures 12–17. In these contour plots onlythe outer 50% span is shown for the uniform grid and the two
Spanwise distributions of circumferentially mass averaged total pressure coefficients
Spanwise distributions of circumferentially mass averaged yaw angles
Legend:
Figure 11. Comparison of spanwise distributions ofcircumferentially mass-averaged Cpt and yaw angles on a fineuniform grid and grids of adaptive test case TC1 and adaptive testcase TC2.
Uniform grid 23 x 60 Total pressure coefficient Cpt
Figure 12. Upper 50% span of Cpt distribution over one pitchmeasured on the uniform grid 23×60.
test cases, TC1 and TC2. The total pressure coefficients areshown in figures 12–14 and the yaw angles in figures 15–17.
2604
A flow adaptive aerodynamic probe concept for turbomachinery
Flow adaptive probe concept Total pressure coefficient Cpt, TC1
Figure 13. Upper 50% span of Cpt distribution over one pitchmeasured on the flow-adapted grid, TC1.
Flow adaptive probe concept Total pressure coefficient Cpt, TC2
Figure 14. Upper 50% span of Cpt distribution over one pitchmeasured on the flow-adapted grid, TC2.
Uniform grid 23 x 60 Yaw angle α
Figure 15. Upper 50% span of α (deg) distribution over one pitchmeasured on the uniform grid 23 × 60.
Overall it can be seen that there is very good qualitativeagreement between the two flow-adapted grid test cases andthe reference case. As described above, more highly refinedmeasurements are obtained at around 60% span in the vicinityof the wake, and around 90% span in the region of the tipleakage flow. Although there is a high density of measurement
Flow adaptive probe concept Yaw angle α, TC1
Figure 16. Upper 50% span of α (deg) distribution over one pitchmeasured on the flow-adapted grid, TC1.
Flow adaptive probe concept Yaw angle α, TC2
Figure 17. Upper 50% span of α (deg) distribution over one pitchmeasured on the flow-adapted grid, TC2.
points in these two regions, a comparison of for examplefigure 16 (test case TC1) and figure 17 (test case TC2) showsthat there are measurement points distributed over the wholemeasurement plane. The user-specified minimum spacingdistance assures that additional points are not inserted in theregions of interest, even if the number of measurement loops ishigh. The small differences between the flow-adapted results,figures 13, 14, 16 and 17, compared to the reference case,figures 12 and 15, are thought to be a result of the non-uniformgrid in the flow-adapted test cases and the uniform grid in thereference case. The measurements on the non-uniform gridhave to be interpolated back onto a uniform grid during thedata processing step described above. It is thought that theuse of a higher order interpolation scheme would minimizethese differences. Nevertheless, the spanwise distributions ofthe circumferentially mass-averaged total pressure coefficientsand yaw angles, shown in figure 11, show that the differencesbetween the flow-adapted grid and uniform grid are small.
The measurement times for the three cases are 8 h for theuniform grid, 1.5 h for test case TC1 and 3 h for test caseTC2. The difference in the measurement times on the flow-adapted grids is a result of the difference in the number ofmeasurement points, 354 points for test case TC1 compared
2605
C Lenherr et al
Total pressure coefficient Cpt
Initial uniform grid PM PG YA
NA =88; NA =206; NA =270; NA =354;
Yaw angle αInitial uniform grid PM PG YA
NA =88; NA =206; NA =270; NA =354;
Figure 18. Evolution of Cpt and α (deg) distributions over one pitch during a run. Test case TC1.
to 513 points for test case TC2. Test case TC1 thus has a gainof 81% in measurement time compared to the uniform grid ofthe reference case.
The evolution of the grid adaptation during a series ofmeasurements is shown in figure 18. Test case TC1 is shownas a representative example. In figure 18 the measuredflowfield on the initial uniform grid, and then at the end ofthe three successive main processing phases, PM, PG andYA, are shown. The corresponding evolution of the mass-averaged total pressure coefficient and yaw angle is shown infigures 19 and 20. An error bar shows the uncertainty in thefive-hole probe measurements and the result from the uniformgrid is shown as a horizontal dashed line that is denoted asthe trendline. It can be seen that 300 measurement pointson the flow-adapted grid yield the same result (within themeasurement uncertainty) as on the uniform grid that has1380 points.
4. Further applications of the method
The applications of the novel flow adaptive algorithm are notlimited to the test case considered above. Below we brieflysummarize two other related applications.
4.1. Fast response aerodynamic probe (FRAP)
An accurate assessment of the losses in turbomachinesrequires the measurement of the unsteady flow field [20–22].These measured flow properties can then be used to derive
Trend for changes in mass averaged Cpt
Figure 19. Evolution of the mass-averaged total pressure coefficientCpt.
parameters such as the non-deterministic pressure coefficientand turbulence intensity that are used to quantify the losses.The initial development of the novel flow adaptive algorithmused the FRAP from which the aforementioned parameterscan be determined. It is thus evident that the novel flowadaptive algorithm has the potential to facilitate the designand development of new turbomachines.
4.2. 1D flow adaptive traversing algorithm
The novel flow adaptive algorithm has been applied to a1D radial traverse. In this case, the algorithm was used to
2606
A flow adaptive aerodynamic probe concept for turbomachinery
Trend for changes in mass averaged yaw angle
Figure 20. Evolution of the mass-averaged yaw angle α (deg).
automatically traverse a probe step-by-step from tip to hub.At each radial position, the previously acquired yaw angle α
is used as an input for the following traverse point. After atraverse, points of interest are determined, and arrays of locallyrefined radial positions are generated. The measurement andrefinement steps are repeated until the algorithm terminatesdue to the limitation criteria. Compared to measurement on astandard uniform grid without refinement, the 1D algorithm is30% faster. As a 1D traverse is made, the detection functionsin the circumferential direction are not evaluated. Therefore,the flow adaptive algorithm can be applied in test facilities inwhich an automated traverse in the circumferential direction isnot available, since each radial traverse can be independentlymade.
5. Conclusions
A novel flow adaptive 2D traversing algorithm has beendeveloped. Measurements in a large axial turbine facilitydemonstrate the potential of the algorithm to substantiallyreduce measurement time, whilst maintaining measurementaccuracy. This novel approach can be used for pneumaticprobe measurements, as well as for unsteady measurementsusing a fast response aerodynamic probe.
The flow adaptive algorithm is comprised of threesequential steps—pre-processing, main processing and post-processing. In the first step an initial uniform grid ofpoints is defined and the flow is measured. Then in thenext step, in an automated manner, the algorithm detectspoints/areas of interest, and adds additional measurementpoints as required. The adaptation is accomplished in boththe radial and circumferential directions in the measurementplane. The detection functions employed in the present workidentify the wakes, which are shed from the rotor and statorblades, and the vortices, that are generated by the passage andleakage flows. In the final step, the measurements on the flow-adapted grid are interpolated back onto a uniform grid in orderto facilitate the post-processing.
The application of the flow adaptive 2D traversingalgorithm shows that compared to a uniform measurementgrid, a flow-adapted grid with 75% fewer measurement points
can be used to resolve the flow field. This reductionin measurement points has a significant impact on themeasurement time; specifically measurement time using theflow adaptive algorithm is 81% quicker than on the uniformmeasurement grid. This measurement time is expected to beeven further reduced in a 3D adaptive flow concept that ispresently under development.
Acknowledgments
The authors would like to express their gratitude tothe technical support personnel of the TurbomachineryLaboratory, and in particular to Cornel Reshef for his help andadvice in the numerous electronic engineering issues that aroseduring the course of the algorithm development. Thanks arealso due to Thomas Behr and Luca Porreca for the installationand the operation of the axial turbine research facility LISA.
References
[1] Sieverding C H, Arts T, Denos R and Brouckaert J F 2000Measurement techniques for unsteady flows inturbomachines Exp. Fluids 28 285–321
[2] Franken A R C and Ivey P C 2002 Reducing aero-enginedevelopment time and costs by using artificial intelligence22nd AIAA Aerodynamic Measurement Technology andGround Testing Conf. (St. Louis, Missouri) AIAA Paper2002-3251
[3] Lenherr C, Oschwald M, Kalfas A I and Abhari R S 2004Flow adaptive aerodynamic probe for turbomachinery flows17th Symp. on Measuring Techniques in Transonic andSupersonic Flows in Cascades and Turbomachines(Stockholm, Sweden)
[4] Behr T, Porreca L, Mokulys T, Kalfas A I and Abhari R S2004 Multistage aspects and unsteady effects of stator androtor clocking in an axial turbine with low aspect ratioblading ASME Turbo Expo (Vienna) ASME Paper GT2004-53612:
[5] Thompson J F, Soni B K and Weatherill N P 1998 Handbookof Grid Generation (Boca Raton, FL: CRC Press)
[6] Kalfas A I and Elder R L 1995 Determination of theintermittency distribution in the boundary layer of a flatplate with c4 leading edge ERCOFTAC Bul. Transit.24 65–7
[7] Binder A and Romey R 1983 Secondary flow effects andmixing of the wake behind a turbine stator ASMEJ. Turbomach. 105 40–6
[8] Treiber M, Kupferschmied P and Gyarmathy G 1998 Analysisof error propagation arising from measurements with aminiature pneumatic 5-hole probe Proc. 14th Symp. onMeasuring Techniques for Transonic and Supersonic Flowsin Cascades and Turbomachines (Limerick, Ireland)
[9] Schlienger J, Pfau A, Kalfas A I and Abhari R S 2002 Singlepressure transducer probe for 3D flow measurements The16th Symp. on Measuring Techniques in Transonic andSupersonic Flow in Cascades and Turbomachines(Cambridge, UK)
[10] Gossweiler C R, Kupferschmied P and Gyarmathy G 1995 Onfast response probes: part 1. Technology, calibration andapplication to turbomachinery ASME J. Turbomach.117 611–7
[11] Johansen E S and Rediniotis O K 2000 Development ofunsteady calibration facilities and techniques forfast-response pressure probes 40th AIAA AerospaceSciences Meeting and Exhib. (Reno, NV) AIAA paper2002-0689
2607
C Lenherr et al
[12] Treaster A L and Yocum A M 1997 The calibration andapplication of five-hole probes ISA Trans. 18 23–34
[13] Pisasale A J and Ahmed N A 2002 Theoretical calibration forhighly three-dimensional low-speed flows of a five-holeprobe Meas. Sci. Technol. 13 1100–7
[14] Humm H, Gossweiler C R and Gyarmathy G 1994 Onfast-response probes: part 2. Aerodynamic design studiesASME J. Turbomach. 117 618–24
[15] Kupferschmied P, Koppel P, Gizzi W, Roduner C andGyarmathy G 2000 Time-resolved flow measurements withfast-response aerodynamic probes in turbomachines Meas.Sci. Technol. 11 1036–54
[16] Kupferschmied P, Koppel P, Roduner C and Gyarmathy G2000 On the development and application of thefast-response aerodynamic probe system in turbomachines:part 1. The measurement system ASME J. Turbomach.122 505–16
[17] Persico G, Gaetani P and Guardone A 2005 Design andanalysis of new concept fast-response pressure probesMeas. Sci. Technol. 16 1741–50
[18] Tanaka K, Kalfas A I and Hodson H P 2000 Development ofsingle sensor fast response pressure probes 15th Symp. onMeasuring Techniques in Transonic and Supersonic Flowsin Cascades and Turbomachines (Florence, Italy)
[19] Schlienger J, Kalfas A I and Abhari R S 2005Vortex-wake-blade interaction in a shrouded axial turbineASME J. Turbomach. 127 699–707
[20] Denton J D 1993 Loss mechanisms in turbomachines ASMEJ. Turbomach. 115 621–56
[21] Sharma O P, Picket G F and Ni R H 1992 Assessment ofunsteady flows in turbines ASME J. Turbomach.114 79–90
[22] Doorly D J 1988 Modelling the unsteady flow in a turbinepassage ASME J. Turbomach. 110 27–37
2608